Author(s): Tae-Woong Kim; Hosung Ahn; Jae-Hyun Ahn; Sung-Ho Byeon; Sung-Wook Wi; Moonil Kim
Linked Author(s):
Keywords: No Keywords
Abstract: Missing data in daily rainfall records need to be filled in accurately beforehand. Presented herein is an effort to develop a new spatial daily rainfall model that is intended specifically to fill in gaps in the measured rainfalls. This study adopted a neural networkoriented pattern classifier that determines a daily rainfall occurrence. We herein tested four alternative classifiers. The testing results reveal that a probabilistic neural network approach was superior to the others. Also, a stepwise regression performed better for estimating rainfall amounts than other competing approaches. This study proved that the proposed model produced accurate and unbiased estimates for missing values of daily rainfall.
Year: 2007